Evolution Towards 5G Multi-tier Cellular Wireless Networks: An Interference Management Perspective

1.1 MB·

by  E. Hossain

1
Evolution Towards 5G Multi-tier Cellular Wireless
Networks: An Interference Management Perspective
Ekram Hossain, Mehdi Rasti, Hina Tabassum, and Amr Abdelnasser
Abstract—The evolving fifth generation (5G) cellular wireless datarateoftheorderof10Gbpswillbeakeyattributeof5G
networksareenvisionedtoovercomethefundamentalchallenges networks.
ofexistingcellularnetworks,e.g.,higherdatarates,excellentend-
5G wireless networks are expected to be a mixture of
to-endperformanceanduser-coverageinhot-spotsandcrowded
network tiers of different sizes, transmit powers, backhaul
areas with lower latency, energy consumption and cost per
informationtransfer.Toaddressthesechallenges,5Gsystemswill connections, different radio access technologies (RATs) that
4 adoptamulti-tierarchitectureconsistingofmacrocells,different are accessed by an unprecedented numbers of smart and het-
1 types of licensed small cells, relays, and device-to-device (D2D) erogeneous wireless devices. This architectural enhancement
0 networks to serve users with different quality-of-service (QoS)
along with the advanced physical communications technol-
2 requirementsinaspectrumandenergy-efficientmanner.Starting
b with the visions and requirements of 5G multi-tier networks, ogy such as high-order spatial multiplexing multiple-input
e this article outlines the challenges of interference management multiple-output (MIMO) communications will provide higher
F (e.g., power control, cell association) in these networks with aggregate capacity for more simultaneous users, or higher
shared spectrum access (i.e., when the different network tiers level spectral efficiency, when compared to the 4G networks.
7 share the same licensed spectrum). It is argued that the existing
Radio resource and interference management will be a key
1 interference management schemes will not be able to address
the interference management problem in prioritized 5G multi- researchchallengeinmulti-tierandheterogeneous5Gcellular
] tier networks where users in different tiers have different networks.Thetraditionalmethodsforradioresourceandinter-
I
N priorities for channel access. In this context, a survey and ference management (e.g., channel allocation, power control,
qualitativecomparisonoftheexistingcellassociationandpower cell association or load balancing) in single-tier networks
.
s control schemes is provided to demonstrate their limitations for (even some of those developed for two-tier networks) may
c interference management in 5G networks. Open challenges are
[ highlighted and guidelines are provided to modify the existing not be efficient in this environment and a new look into the

schemes in order to overcome these limitations and make them interference management problem will be required.
2
suitable for the emerging 5G systems. First, the article outlines the visions and requirements of
v
0 Index Terms—5G cellular wireless, multi-tier networks, inter- 5G cellular wireless systems. Major research challenges are
3 ference management, cell association, power control. then highlighted from the perspective of interference manage-
5 ment when the different network tiers share the same radio
5
spectrum. A comparative analysis of the existing approaches
. I. INTRODUCTION
1 for distributed cell association and power control (CAPC) is
0 Tosatisfytheever-increasingdemandformobilebroadband thenprovidedfollowedbyadiscussionontheirlimitationsfor
4 communications, the IMT-Advanced (IMT-A) standards have
5G multi-tier cellular networks. Finally, a number of sugges-
1 been ratified by the International Telecommunications Union
tions are provided to modify the existing CAPC schemes to
:
v (ITU) in November 2010 and the fourth generation (4G) overcome these limitations.
i wireless communication systems are currently being deployed
X
worldwide.ThestandardizationforLTERel-12,alsoknownas
ar LTE-B, is also ongoing and expected to be finalized in 2014. II. VISIONSANDREQUIREMENTSFOR5GMULTI-TIER
Nonetheless,existingwirelesssystemswillnotbeabletodeal CELLULARNETWORKS
with the thousand-fold increase in total mobile broadband 5Gmobileandwirelesscommunicationsystemswillrequire
data [1] contributed by new applications and services such amixofnewsystemconceptstoboostthespectralandenergy
as pervasive 3D multimedia, HDTV, VoIP, gaming, e-Health, efficiency. The visions and requirements for 5G wireless
andCar2xcommunication.Inthiscontext,thefifthgeneration systems are outlined below.
(5G) wireless communication technologies are expected to
• Data rate and latency: For dense urban areas, 5G net-
attain 1000 times higher mobile data volume per unit area,
works are envisioned to enable an experienced data rate
10-100 times higher number of connecting devices and user
of 300 Mbps and 60 Mbps in downlink and uplink,
data rate, 10 times longer battery life and 5 times reduced
respectively, in 95% of locations and time [2]. The end-
latency [2]. While for 4G networks the single-user average
to-end latencies are expected to be in the order of 2 to
data rate is expected to be 1 Gbps, it is postulated that cell
5 milliseconds. The detailed requirements for different
scenarios are listed in [2].
E. Hossain, H. Tabassum, and A. Abdelnasser are with the Department
of Electrical and Computer Engineering at the University of Manitoba, • Machine-type Communication (MTC) devices: The num-
Canada. M. Rasti is with the Department of Computer Engineering and ber of traditional human-centric wireless devices with
InformationTechnology,AmirkabirUniversityofTechnology,Iran.Thiswork
Internet connectivity (e.g., smart phones, super-phones,
wassupportedbytheNaturalSciencesandEngineeringResearchCouncilof
Canada(NSERC)StrategicProjectGrantSTPGP430285-12. tablets) may be outnumbered by MTC devices which
2
can be used in vehicles, home appliances, surveillance D2D communications, where the macrocell BS performs
devices, and sensors. control signaling in terms of synchronization, beacon
• Millimeter-wave communication: To satisfy the expo- signal configuration and providing identity and security
nential increase in traffic and the addition of different management[3].Thisfeaturewillextendin5Gnetworks
devices and services, additional spectrum beyond what to allow other nodes, rather than the macrocell BS, to
waspreviouslyallocatedto4Gstandardissoughtfor.The havethecontrol.Forexample,consideraD2Dlinkatthe
use of millimeter-wave frequency bands (e.g., 28 GHz celledgeandthedirectlinkbetweentheD2Dtransmitter
and 38 GHz bands) is a potential candidate to overcome UE to the macrocell is in deep fade, then the relay node
the problem of scarce spectrum resources since it allows can be responsible for the control signaling of the D2D
transmission at wider bandwidths than conventional 20 link (i.e., relay-aided D2D communication).
MHz channels for 4G systems. • Energy harvesting for energy-efficient communication:
• Multiple RATs: 5G is not about replacing the existing Oneofthemainchallengesin5Gwirelessnetworksisto
technologies, but it is about enhancing and support- improve the energy efficiency of the battery-constrained
ing them with new technologies [1]. In 5G systems, wireless devices. To prolong the battery lifetime as well
the existing RATs, including GSM (Global System for as to improve the energy efficiency, an appealing so-
Mobile Communications), HSPA+ (Evolved High-Speed lution is to harvest energy from environmental energy
Packet Access), and LTE, will continue to evolve to sources (e.g., solar and wind energy). Also, energy can
provide a superior system performance. They will also be harvested from ambient radio signals (i.e., RF en-
be accompanied by some new technologies (e.g., beyond ergy harvesting) with reasonable efficiency over small
LTE-Advanced). distances. The havested energy could be used for D2D
• Base station (BS) densification: BS densification is an communication or communication within a small cell.
effective methodology to meet the requirements of 5G In this context, simultaneous wireless information and
wirelessnetworks.Specifically,in5Gnetworks,therewill power transfer (SWIPT) is a promising technology for
be deployments of a large number of low power nodes, 5Gwirelessnetworks.However,practicalcircuitsforhar-
relays, and device-to-device (D2D) communication links vestingenergyarenotyetavailablesincetheconventional
with much higher density than today’s macrocell net- receiver architecture is designed for information transfer
works. Fig. 1 shows such a multi-tier network with a only and, thus, may not be optimal for SWIPT. This is
macrocell overlaid by relays, picocells, femtocells, and due to the fact that both information and power transfer
D2D links. The adoption of multiple tiers in the cellular operate with different power sensitivities at the receiver
network architecture will result in better performance in (e.g., -10dBm and -60dBm for energy and information
terms of capacity, coverage, spectral efficiency, and total receivers, respectively) [4]. Also, due to the potentially
powerconsumption,providedthattheinter-tierandintra- low efficiency of energy harvesting from ambient radio
tier interferences are well managed. signals,acombinationofdifferentenergyharvestingtech-
• Prioritized spectrum access: The notions of both traffic- nologies may be required for macrocell communication.
based and tier-based priorities will exist in 5G networks.
Traffic-based priority arises from the different require-
III. INTERFERENCEMANAGEMENTCHALLENGESIN5G
ments of the users (e.g., reliability and latency require-
MULTI-TIERNETWORKS
ments,energyconstraints),whereasthetier-basedpriority
is for users belonging to different network tiers. For The key challenges for interference management in 5G
example, with shared spectrum access among macrocells multi-tier networks will arise due to the following reasons
and femtocells in a two-tier network, femtocells create which affect the interference dynamics in the uplink and
“dead zones” around them in the downlink for macro downlink of the network: (i) heterogeneity and dense de-
users. Protection should, thus, be guaranteed for the ployment of wireless devices, (ii) coverage and traffic load
macro users. Consequently, the macro and femtousers imbalance due to varying transmit powers of different BSs
play the role of high-priority users (HPUEs) and low- in the downlink, (iii) public or private access restrictions in
priority users (LPUEs), respectively. In the uplink di- different tiers that lead to diverse interference levels, and (iv)
rection, the macrocell users at the cell edge typically the priorities in accessing channels of different frequencies
transmit with high powers which generates high uplink and resource allocation strategies. Moreover, the introduction
interference to nearby femtocells. Therefore, in this case, of carrier aggregation, cooperation among BSs (e.g., by using
the user priorities should get reversed. Another example coordinatedmulti-pointtransmission(CoMP))aswellasdirect
is a D2D transmission where different devices may op- communication among users (e.g., D2D communication) may
portunisticallyaccessthespectrumtoestablishacommu- furthercomplicatethedynamicsoftheinterference.Theabove
nicationlinkbetweenthemprovidedthattheinterference factors translate into the following key challenges.
introduced to the cellular users remains below a given • Designing optimized cell association and power control
threshold. In this case, the D2D users play the role of (CAPC) methods for multi-tier networks: Optimizing the
LPUEswhereasthecellularusersplaytheroleofHPUEs. cell associations and transmit powers of users in the
• Network-assisted D2D communication: In the LTE Rel- uplink or the transmit powers of BSs in the downlink
12 and beyond, focus will be on network controlled are classical techniques to simultaneously enhance the
3
UE Relay Mobile Core
Network
UE
D2D
Femtocell
D2D UE User
Broadband
Femtocell Internet Connection
Picocell
Picocell
Macrocell
UE
Relay
Relay
D2D
D2D UE
Femtocell
Picocell
Fig.1. Amulti-tiernetworkcomposedofmacrocells,picocells,femtocells,relays,andD2Dlinks.Arrowsindicatewirelesslinks,whereasthedashedlines
denotethebackhaulconnections.
system performance in various aspects such as interfer- sociation to multiple BSs: Compared to existing CAPC
ence mitigation, throughput maximization, and reduction schemes in which each user can associate to a single
in power consumption. Typically, the former is needed BS, simultaneous connectivity to several BSs could be
to maximize spectral efficiency, whereas the latter is possible in 5G multi-tier network. This would enhance
required to minimize the power (and hence minimize the the system throughput and reduce the outage ratio by
interferencetootherlinks)whilekeepingthedesiredlink effectively utilizing the available resources, particularly
quality. Since it is not efficient to connect to a congested for cell edge users. Thus the existing CAPC schemes
BS despite its high achieved signal-to-interference ratio should be extended to efficiently support simultaneous
(SIR), cell association should also consider the status associationofausertomultipleBSsanddetermineunder
of each BS (load) and the channel state of each UE. which conditions a given UE is associated to which BSs
The increase in the number of available BSs along with in the uplink and/or downlink.
multi-pointtransmissionsandcarrieraggregationprovide • Designing efficient methods for cooperation and coor-
multiple degrees of freedom for resource allocation and dination among multiple tiers: Cooperation and coordi-
cell-selectionstrategies.Forpowercontrol,thepriorityof nation among different tiers will be a key requirement
different tiers need also be maintained by incorporating to mitigate interference in 5G networks. Cooperation
the quality constraints of HPUEs. between the macrocell and small cells was proposed for
Unlike downlink, the transmission power in the uplink LTE Rel-12 in the context of soft cell, where the UEs
depends on the user’s battery power irrespective of the are allowed to have dual connectivity by simultaneously
type of BS with which users are connected. The battery connecting to the macrocell and the small cell for uplink
powerdoesnotvarysignificantlyfromusertouser;there- and downlink communications or vice versa [3]. As
fore,theproblemsofcoverageandtrafficloadimbalance has been mentioned before in the context of asymmetry
may not exist in the uplink. This leads to considerable of transmission power in uplink and downlink, a UE
asymmetries between the uplink and downlink user as- mayexperiencethehighestdownlinkpowertransmission
sociation policies. Consequently, the optimal solutions from the macrocell, whereas the highest uplink path
for downlink CAPC problems may not be optimal for gain may be from a nearby small cell. In this case,
the uplink. It is therefore necessary to develop joint the UE can associate to the macrocell in the downlink
optimization frameworks that can provide near-optimal, and to the small cell in the uplink. CoMP schemes
if not optimal, solutions for both uplink and downlink. based on cooperation among BSs in different tiers (e.g.,
Moreover, to deal with this issue of asymmetry, separate cooperation between macrocells and small cells) can be
uplink and downlink optimal solutions are also useful as developed to mitigate interference in the network. Such
far as mobile users can connect with two different BSs schemes need to be adaptive and consider user locations
for uplink and downlink transmissions which is expected as well as channel conditions to maximize the spectral
to be the case in 5G multi-tier cellular networks [3]. and energy efficiency of the network. This cooperation
• Designing efficient methods to support simultaneous as- however, requires tight integration of low power nodes
4
into the network through the use of reliable, fast and onalization in which specific sub-frames are left blank
low latency backhaul connections which will be a major by the unbiased BS and off-loaded users are scheduled
technical issue for upcoming multi-tier 5G networks. within these sub-frames to avoid inter-tier interference.
Intheremainingofthisarticle,wewillfocusonthereview This improves the overall throughput of the off-loaded
of existing power control and cell association strategies to users by sacrificing the time sub-frames and throughput
demonstrate their limitations for interference management in oftheunbiasedBS.Thelargerbiasvaluesresultinhigher
5Gmulti-tierprioritizedcellularnetworks(i.e.,whereusersin degree of offloading and thus require more blank sub-
different tiers have different priorities depending on the loca- frames to protect the offloaded users. Given a specific
tion, application requirements and so on). Design guidelines number of ABSs or the ratio of blank over total number
will then be provided to overcome these limitations. Note that of sub-frames (i.e., ABS ratio) that ensures the minimum
issues such as channel scheduling in frequency domain, time- throughput of the unbiased BSs, this criterion allows
domaininterferencecoordinationtechniques(e.g.,basedonal- a user to select a cell with maximum ABS ratio and
most blank subframes), coordinated multi-point transmission, may even associate with the unbiased BS if ABS ratio
and spatial domain techniques (e.g., based on smart antenna decreases significantly.
techniques) are not considered in this article. A qualitative comparison among these cell association
schemes is given in Table I. The specific key terms used in
IV. DISTRIBUTEDCELLASSOCIATIONANDPOWER TableIaredefinedasfollows:channel-awareschemesdepend
CONTROLSCHEMES:CURRENTSTATEOFTHEART ontheknowledgeofinstantaneouschannelandtransmitpower
atthereceiver.Theinterference-awareschemesdependonthe
A. Distributed Cell Association Schemes
knowledge of instantaneous interference at the receiver. The
The state-of-the-art cell association schemes that are cur- load-aware schemes depend on the traffic load information
rently under investigation for multi-tier cellular networks are (e.g., number of users). The resource-aware schemes require
reviewed and their limitations are explained below. the resource allocation information (i.e., the chance of getting
• Reference Signal Received Power (RSRP)-based scheme a channel or the proportion of resources available in a cell).
[5]: A user is associated with the BS whose signal is The priority-aware schemes require the information regarding
received with the largest average strength. A variant of thepriorityofdifferenttiersandallowaprotectiontoHPUEs.
RSRP, i.e., Reference Signal Received Quality (RSRQ) All of the above mentioned schemes are independent, dis-
isalsousedforcellselectioninLTEsingle-tiernetworks tributed, andcan beincorporated with anytype of powercon-
which is similar to the signal-to-interference (SIR)-based trol scheme. Although simple and tractable, the standard cell
cell selection where a user selects a BS communicating association schemes, i.e., RSRP, RSRQ, and CRE are unable
with which gives the highest SIR. In single-tier networks to guarantee the optimum performance in multi-tier networks
with uniform traffic, such a criterion may maximize the unless critical parameters, such as bias values, transmit power
network throughput. However, due to varying transmit of the users in the uplink and BSs in the downlink, resource
powers of different BSs in the downlink of multi-tier partitioning, etc. are optimized.
networks, such cell association policies can create a
B. Distributed Power Control Schemes
huge traffic load imbalance. This phenomenon leads to
overloading of high power tiers while leaving low power From a user’s point of view, the objective of power control
tiers underutilized. is to support a user with its minimum acceptable throughput,
• Bias-based Cell Range Expansion (CRE) [6]: The idea whereas from a system’s point of view it is to maximize
of CRE has been emerged as a remedy to the problem the aggregate throughput. In the former case, it is required
of load imbalance in the downlink. It aims to increase to compensate for the near-far effect by allocating higher
the downlink coverage footprint of low power BSs by power levels to users with poor channels as compared to
addingapositivebiastotheirsignalstrengths(i.e.,RSRP UEs with good channels. In the latter case, high power levels
or RSRQ). Such BSs are referred to as biased BSs. are allocated to users with best channels and very low (even
This biasing allows more users to associate with low zero) power levels are allocated to others. The aggregate
power or biased BSs and thereby achieve a better cell transmitpower,theoutageratio,andtheaggregatethroughput
load balancing. Nevertheless, such off-loaded users may (i.e., the sum of achievable rates by the UEs) are the most
experience unfavorable channel from the biased BSs and important measures to compare the performance of different
strong interference from the unbiased high-power BSs. power control schemes. The outage ratio of a particular tier
The trade-off between cell load balancing and system can be expressed as the ratio of the number of UEs supported
throughputthereforestrictlydependsontheselectedbias byatierwiththeirminimumtargetSIRsandthetotalnumber
values which need to be optimized in order to maximize of UEs in that tier.
the system utility. In this context, a baseline approach in Numerouspowercontrolschemeshavebeenproposedinthe
LTE-Advancedisto“orthogonalize”thetransmissionsof literature for single-tier cellular wireless networks. According
the biased and unbiased BSs in time/frequency domain to the corresponding objective functions and assumptions, the
such that an interference-free zone is created. schemes can be classified into the following four types.
• Association based on Almost Blank Sub-frame (ABS) • Target-SIR-trackingpowercontrol(TPC)[8]:IntheTPC,
ratio [7]: The ABS technique uses time domain orthog- each UE tracks its own predefined fixed target-SIR.
5
TABLEI
QUALITATIVECOMPARISONOFEXISTINGCELLASSOCIATIONSCHEMESFORMULTI-TIERNETWORKS
RSRP[5] RSRQ[5] CRE[6] ABSratio[7]
Objective Maximizereceivedsignal MaximizeSIR Balancetrafficload Maximize rate and
power balancetrafficload
Applicability Uplinkanddownlink Uplinkanddownlink Downlink Downlink
Channel-aware (cid:88) (cid:88) (cid:88) (cid:88)
Interference-aware X (cid:88) (cid:88) (cid:88)
Trafficload-aware X X (cid:88) (cid:88)
Resource-aware X X X (cid:88)
Priority-aware X X X (cid:88)
The TPC enables the UEs to achieve their fixed target- power consumption) as compared to TPC.
SIRsatminimalaggregatetransmitpower,assumingthat
The aforementioned state-of-the-art distributed power con-
the target-SIRs are feasible. However, when the system
trol schemes for satisfying various objectives in single-tier
is infeasible, all non-supported UEs (those who can-
wireless cellular networks are unable to address the interfer-
not obtain their target-SIRs) transmit at their maximum
ence management problem in prioritized 5G multi-tier net-
power,whichcausesunnecessarypowerconsumptionand
works.Thisisduetothefactthattheydonotguaranteethatthe
interference to other users, and therefore, increases the
total interference caused by the LPUEs to the HPUEs remain
number of non-supported UEs.
within tolerable limits, which can lead to the SIR outage of
• TPCwithgradualremoval(TPC-GR)[9],[10],and[11]: some HPUEs. Thus there is a need to modify the existing
To decrease the outage ratio of the TPC in an infeasible
schemessuchthatLPUEstracktheirobjectiveswhilelimiting
system, a number of TPC-GR algorithms were proposed
theirtransmitpowertomaintainagiveninterferencethreshold
inwhichnon-supportedusersreducetheirtransmitpower
at HPUEs.
[10] or are gradually removed [9], [11].
A qualitative comparison among various state-of-the-art
• Opportunistic power control (OPC) [12]: From the sys- power control problems with different objectives and con-
tem’s point of view, OPC allocates high power levels to
straints and their corresponding existing distributed solutions
users with good channels (experiencing high path-gains
areshowninTableII.Thistablealsoshowshowtheseschemes
and low interference levels) and very low power to users
canbemodifiedandgeneralizedfordesigningCAPCschemes
with poor channels. In this algorithm, a small difference
for prioritized 5G multi-tier networks.
in path-gains between two users may lead to a large
differenceintheiractualthroughputs[12].OPCimproves
the system performance at the cost of reduced fairness C. Joint Cell Association and Power Control Schemes
among users.
Avery fewwork intheliterature haveconsidered theprob-
• Dynamic-SIRtrackingpowercontrol(DTPC)[13]:When lem of distributed CAPC jointly (e.g., [14]) with guaranteed
the target-SIR requirements for users are feasible, TPC
convergence.Forsingle-tiernetworks,adistributedframework
causes users to exactly hit their fixed target-SIRs even
for uplink was developed [14], which performs cell selection
if additional resources are still available that can other-
based on the effective-interference (ratio of instantaneous
wise be used to achieve higher SIRs (and thus better
interference to channel gain) at the BSs and minimizes the
throughputs). Besides, the fixed-target-SIR assignment
aggregateuplinktransmitpowerwhileattainingusers’desired
is suitable only for voice service for which reaching a
SIR targets. Following this approach, a unified distributed
SIR value higher than the given target value does not
algorithmwasdesignedin[15]fortwo-tiernetworks.Thecell
affect the service quality significantly. In contrast, for
association is based on the effective-interference metric and is
data services, a higher SIR results in a better throughput,
integrated with a hybrid power control (HPC) scheme which
which is desirable. The DTPC algorithm was proposed
is a combination of TPC and OPC power control algorithms.
in [13] to address the problem of system throughput
Although the above frameworks are distributed and opti-
maximization subject to a given feasible lower bound for
mal/suboptimal with guaranteed convergence in conventional
the achieved SIRs of all users in cellular networks. In
networks, they may not be directly compatible to the 5G
DTPC,eachuserdynamicallysetsitstarget-SIRbyusing
multi-tier networks. The interference dynamics in multi-tier
TPC and OPC in a selective manner. It was shown that
networksdependssignificantlyonthechannelaccessprotocols
when the minimum acceptable target-SIRs are feasible,
(or scheduling), QoS requirements and priorities at different
the actual SIRs received by some users can be dynam-
tiers. Thus, the existing CAPC optimization problems should
ically increased (to a value higher than their minimum
be modified to include various types of cell selection methods
acceptable target-SIRs) in a distributed manner so far
(some examples are provided in Table I) and power control
as the required resources are available and the system
methods with different objectives and interference constraints
remains feasible (meaning that reaching the minimum
(e.g.,interferenceconstraintsformacrocellUEs,picocellUEs,
target-SIRs for the remaining users are guaranteed). This
or D2D receiver UEs). A qualitative comparison among the
enhances the system throughput (at the cost of higher
existing CAPC schemes along with the open research areas
6
are highlighted in Table II. A discussion on how these open high-priority macro cell) ranging from 3 to 6 with step size
problems can be addressed is provided in the next section. of1(i.e.,n=3,4,5,6),averagedover100independentsnap-
shotsforauniformdistributionofBSsandusers’locationsas
V. DESIGNGUIDELINESFORDISTRIBUTEDCAPC explained above. Fig. 2(a) illustrates the impact of employing
SCHEMESIN5GMULTI-TIERNETWORKS either of existing distributed algorithms TPC and TPC-GR
employed by LPUEs on the outage of HPUEs (which employ
Interference management in 5G networks requires efficient
TPC).Ascanbeseen,althoughtheoutageratiofortheLPUEs
distributed CAPC schemes such that each user can possibly
and HPUEs are improved by TPC-GR, as compared to TPC,
connect simultaneously to multiple BSs (can be different
protection of the HPUEs is not guaranteed by the TPC and
for uplink and downlink), while achieving load balancing in
TPC-GRalgorithms.Incontrast,ourproposedprioritizedTPC
different cells and guaranteeing interference protection for the
and TPC-GR algorithms guarantee protection of the HPUEs
HPUEs. In what follows, we provide a number of suggestions
at the cost of increased outage ratio for the LPUEs. Similar
to modify the existing schemes.
resultsareachievedwhentheHPUEsrigidlytracktheirtarget-
SIRbyemployingTPCandtheLPUEsemploytheprioritized
A. Prioritized Power Control
OPC. That is, protection of the HPUEs is guaranteed at the
To guarantee interference protection for HPUEs, a possible cost of decreased system throughput for the LPUEs.
strategy is to modify the existing power control schemes
listed in the first column of Table II such that the LPUEs
B. Resource-Aware Cell Association Schemes
limit their transmit power to keep the interference caused
to the HPUEs below a predefined threshold, while tracking Cellassociationschemesneedtobedevisedthatcanbalance
their own objectives. In other words, as long as the HPUEs the traffic load as well as minimize interference or maximize
are protected against existence of LPUEs, the LPUEs could SIRlevelsatthesametimeandcanachieveagoodbalancebe-
employ an existing distributed power control algorithm to tweentheseobjectiveswithouttheneedofstaticbiasing-based
satisfy a predefined goal. This offers some fruitful direction CRE or ABS schemes. As an example, instead of sacrificing
for future research and investigation as stated in Table II. theresourcesofahigh-powerBStoprotecttheoffloadedusers
To address these open problems in a distributed manner, the (e.g., as in CRE and ABS technique as detailed in Section
existing schemes should be modified so that the LPUEs in IV.A), user association schemes can also be developed in
addition to setting their transmit power for tracking their which a user always prefers to associate with a low-power
objectives,limittheirtransmitpowertokeeptheirinterference BS (with no bias) as long as the received interference from
on receivers of HPUEs below a given threshold. This could high-power BS remains below a threshold. The high-power
be implemented by sending a command from HPUEs to its BS may consider minimizing its transmit power subject to
nearby LPUEs (like a closed-loop power control command a maximum interference level experienced by the off-loaded
used to address the near-far problem), when the interference users (i.e., prioritized power control in the downlink).
causedbytheLPUEstotheHPUEsexceedsagiventhreshold. The CRE technique forces the users to select low power
We refer to this type of power control as prioritized power nodesbyaddingafixedbiastothemfortrafficloadbalancing.
control. Note that the notion of priority and thus the need of However, this strategy is immune to the resource allocation
prioritizedpowercontrolexistsimplicitlyindifferentscenarios criterion employed in the corresponding cell. For instance,
of 5G networks, as briefly discussed in Section II. Along this if a low-power BS performs greedy scheduling, it is highly
line, some modified power control optimization problems are unlikely that an off-loaded user will get a channel (i.e., low
formulated for 5G multi-tier networks in second column of channel access probability) even if the RSRP with bias is the
Table II. best towards that BS among all other BSs. For round-robin
To compare the performance of existing distributed power scheduling, if the low-power BS has a large number of users,
control algorithms, let us consider a prioritized multi-tier itmaykeeptheoff-loadedusersinstarvationforlongtimeand
cellular wireless network where a high-priority tier consisting therefore cause delay. Clearly, the channel access probability
of 3×3 macro cells, each of which covers an area of 1000 plays a major role in cell-association methods. Thus, the
m × 1000 m, coexists with a low-priority tier consisting of n bias selection should be adaptive (instead of static) to the
small-cells per each high-priority macro cell, each of which resourceallocationcriterion,trafficload,anddistance/channel
covers an area of 200 m × 200 m within the coverage area corresponding to the different BSs.
of a high-priority large cell. Each user is associated with only In this context, new cell association schemes/metrics need
oneBSofitscorrespondingtier.Eachhigh-priorityBSorlow- to be developed that can optimize multiple objectives, e.g.,
priorityBSislocatedatthecentreofitscorrespondingcelland traffic-loadbalancingandrate-maximizationatthesametime.
serves 5 HPUEs and 4 LPUEs, respectively. The target-SIRs To illustrate this, we introduce a new resource-aware cell
are considered to be the same for all users. Suppose that all association criterion in which each user selects a BS with
LPUEsemployeitherTPC,TPC-GR,ourproposedprioritized maximum channel access probability, i.e., max{p }, where p
i i
TPC or prioritized TPC-GR, while all of the HPUEs rigidly is the channel access probability of a cell i. Note that, the
track their target-SIRs (i.e., HPUEs employ TPC). metric p varies for different resource allocation criteria at
i
Fig.2illustratestheoutageratiofortheHPUEsandLPUEs the BSs. For instance, in round-robin scheduling, p is the
i
versus the total number of low-priority small cells (per each reciprocal of the number of users. On the other hand, for
7
(a) (b)
Fig.2. TheoutageratioforLPUEsandHPUEsversusthetotalnumberoflow-prioritysmallcellspereachhigh-prioritycell,forthefollowingdistributed
powercontrolalgorithms:TPC[8],TPC-GR[10],prioritizedTPC,andprioritizedTPC-GR.
greedy scheduling p is the probability that the channel gain to load-aware cell association for round-robin scheduling) can
i
of a potential admitting user exceeds the channel gain of all degrade the spectral efficiency performance significantly due
existing users in cell i and thus depends on both channel and to strong close-by interferers. On the other hand, a hybrid
number of users in cell i. This new metric implicitly tends to schemecanachievesignificantperformancegainsperchannel
balance the traffic load since if the number of users grows for a typical user, especially in sparse deployments.
in a cell, p reduces and stops any further associations or
i
vice versa. In this way, the proposed criterion p provides
i
an adaptive biasing to different BSs considering their corre-
spondingschedulingscheme,trafficloadandchannelgains(if
opportunistic scheduling is employed). 0.2
Hybrid Cell Association
Note that, in distance-aware cell association, each user Distance−Aware Cell Association
selects a cell with minimum distance which tends to improve 0.18 Resource−Aware Cell Association
the sum-rate performance. However, this criterion is immune ps/Hz]0.16
b
to traffic load conditions. Combining the aforementioned el [
n
resource-aware and distance-aware criteria, we now consider an0.14
h
C
a hybrid cell association. The hybrid cell association scheme er
allows a typical user to select a cell with the maximum of er p0.12
Us
product of distance-based channel gain and pi. If pi =0 (i.e., per 0.1
high/infinitetrafficload),auserwillnotselectcellievenifits ncy
the closest cell and vice versa. Thus, hybrid schemes assist in Efficie0.08
achieving a good balance between traffic-load balancing and al
throughput maximization. ectr0.06
p
S
For demonstration purpose, a quantitative comparison of a 0.04
resource-aware, distance-aware and a hybrid cell association
0.02
scheme is shown in Fig. 3 for round-robin scheduling. We 5 10 15 20 25 30
Number of Small Cells
consider downlink transmission in a multi-tier network with a
circular macrocell overlaid by randomly deployed small cells.
The average spectral efficiency of a macro user on a given Fig.3. Comparisonamongdistance-aware,resource-aware,andhybridcell-
association schemes (for path-loss exponent = 4, macrocell transmit power
channeliscomputedasafunctionofthenumberofsmall-cells
=10Wandsmallcelltransmitpower=1Wforthesimulationsetupfora
in a macrocell. Each small cell has a different user intensity givenmacrocellshowninFig.4).
witharrivalofusersmodeledbyPoissondistribution,thus,cre-
ating a non-uniform traffic load scenario per small cell. It can
be seen that only resource-aware association (which reduces
8
VI. CONCLUSION
600
We have outlined the challenges for interference manage-
mentin5Gmulti-tiernetworksconsideringitsvisions,require-
400 ments,andkeyfeatures. Thesenetworkswillbecharacterized
by the existence of different access priority for users and
200 tiers along with the possibility of simultaneous connectivity
of users to multiple BSs. Along with these features, different
BSassociationforuplinkanddownlinktransmissionopennew
0
challenges and at the same time increase degrees of freedom
for power control and cell association. Open challenges have
−200 beenhighlightedandguidelineshavebeenprovidedtomodify
Macrocell Users the existing schemes in order to make them suitable for 5G
Small cells multi-tier networks. In this context, a promising direction for
−400 Macrocell
Small cell BSs future research is to devise efficient joint CAPC methods
Small cell Users
Macrocell BS that satisfy objectives such as maximizing system throughput,
−600 balance traffic load subject to a minimum SIR for high
−600 −400 −200 0 200 400 600
priority users. To address these multiple objectives, resource-
aware user association can be combined with conventional
Fig. 4. A circular macrocell with several small cells. Each small cell has cell association methods to satisfy the required objectives.
varyingusertrafficload. Thehybridcellassociationmethodscombinedwithprioritized
powercontrolwillbeamongthekeyenablersforevolving5G
cellular networks.
C. Resource-Aware Cell Association and Prioritized Power
Control
REFERENCES
Simultaneous connections to multiple BSs and different
[1] Ericsson,“5GRadioAccess,ResearchandVision,”whitepaper,2013.
BS association for uplink and downlink would increase the [2] Metis, “Scenarios, requirements and KPIs for 5G mobile and wireless
degreesof freedomwhichcanbeexploited tofurtherimprove system,”ICT-317669METISproject,May2013.
[3] D.Astely,E.Dahlman,G.Fodor,S.Parkvall,andJ.Sachs,“LTErelease
thenetworkcapacityandbalancetheloadamongdifferentBSs
12 and beyond [accepted from open call],” IEEE Communications
in different tiers. The existing criteria for cell association can Magazine,vol.51,no.7,pp.154–160,2013.
begeneralizedtosupportsimultaneousconnectiontomultiple [4] X. Zhou, R. Zhang, and C. K. Ho, “Wireless information and power
BSs. For instance, the minimum effective-interference-based transfer:Architecturedesignandrate-energytradeoff,”inProceedingsof
IEEEGlobalCommunicationsConference(GLOBECOM’12),pp.3982–
cell association can be generalized so that when the differ-
3987,Dec.2012.
encesamongeffective-interferencelevelsbetweenagivenuser [5] J. Sangiamwong, Y. Saito, N. Miki, T. Abe, S. Nagata, and Y. Oku-
and some BSs which offer that user the lowest effective- mura,“Investigationoncellselectionmethodsassociatedwithinter-cell
interference coordination in heterogeneous networks for lte-advanced
interference levels is not large, that user can simultaneously
downlink,”inProceedingsofEuropeanWirelessConferenceSustainable
connects to those BSs. The proposed resource-aware criterion WirelessTechnologies,pp.1–6,2011.
for cell association can then be combined with this criterion [6] I. Guvenc, “Capacity and fairness analysis of heterogeneous networks
with range expansion and interference coordination,” IEEE Communi-
to balance the traffic load.
cationsLetters,vol.15,no.10,pp.1084–1087,2011.
These cell-association methods can be combined with the [7] J. Oh and Y. Han, “Cell selection for range expansion with almost
prioritized power control schemes depending on the desired blank subframe in heterogeneous networks,” in Proceedings of IEEE
International Symposium on Personal Indoor and Mobile Radio Com-
objectives.Animportantissueinthisregardistoselectacor-
munications(PIMRC’12),pp.653–657,2012.
rectcombinationofcell-associationandpowercontrolmethod [8] G. Foschini and Z. Miljanic, “A simple distributed autonomous power
to achieve a given objective. For instance, joint minimum controlalgorithmanditsconvergence,”IEEETransactionsonVehicular
Technology,vol.42,no.4,pp.641–646,1993.
effective-interference based cell association and OPC is not
[9] M. Rasti, A.-R. Sharafat, and J. Zander, “Pareto and energy-efficient
capable of addressing the objective of throughput maximiza- distributed power control with feasibility check in wireless networks,”
tion(P3)intheuplink,asinthiscasealluserstrytoassociate IEEETransactionsonInformationTheory,vol.57,no.1,pp.245–255,
2011.
withaBSofminimumeffectiveinterferencewhichultimately
[10] M.RastiandA.-R.Sharafat,“Distributeduplinkpowercontrolwithsoft
results in high transmit power of all users. Although the removalforwirelessnetworks,”IEEETransactionsonCommunications,
system throughput is improved when users with good channel vol.59,no.3,pp.833–843,2011.
[11] F. Berggren, R. Jantti, and S.-L. Kim, “A generalized algorithm for
conditions increase their transmit power, it degrades when
constrainedpowercontrolwithcapabilityoftemporaryremoval,”IEEE
users with poor channel conditions increase their transmit Transactions on Vehicular Technology, vol. 50, no. 6, pp. 1604–1612,
power. Thus, the need to consider a different cell-association 2001.
[12] K.-K.LeungandC.-W.Sung,“Anopportunisticpowercontrolalgorithm
schemetoachievetheobjectivesofprioritizedpowercontrolis
forcellularnetwork,”IEEE/ACMTransactionsonNetworking,vol.14,
evident.InconjunctionwithOPC,itmaybeusefultoconsider no.3,pp.470–478,2006.
RSRP or RSRQ-based cell-association techniques that will [13] M.Rasti,A.-R.Sharafat,andJ.Zander,“Adistributeddynamictarget-
sir-tracking power control algorithm for wireless cellular networks,”
allowcellassociationsbasedontheirchannelconditionsrather
IEEETransactionsonVehicularTechnology,vol.59,no.2,pp.906–916,
than the received interference. 2010.
9
[14] R.D.YatesandC.-Y.Huang,“Integratedpowercontrolandbasestation Hina Tabassum received the B.Eng. degree in
assignment,”IEEETransactionsonVehicularTechnology,vol.44,no.3, electronicengineeringfromtheN.E.DUniversityof
pp.638–644,1995. Engineering and Technology (NEDUET), Karachi,
[15] H.N.VuandL.B.Le,“Distributedbasestationassociationandpower Pakistan,in2004.Duringherundergraduatestudies
control for heterogeneous cellular networks,” IEEE Transactions on PLACE she received the Gold medal from NEDUET and
VehicularTechnology,vol.63,no.1,pp.282–296,2014. PHOTO fromSIEMENSforsecuringthefirstpositionamong
HERE all engineering universities of Karachi. She then
worked as a lecturer in NEDUET for two years.
In September 2005, she joined the Pakistan Space
andUpperAtmosphereResearchCommission(SU-
PARCO), Karachi, Pakistan and there she received
the best performance award in 2009. She completed her Masters and Ph.D.
degreesincommunicationsengineering,respectively,fromNEDUETin2009
andKingAbdullahUniversityofScienceandTechnology(KAUST),Makkah
Province, Saudi Arabia, in May 2013. Currently, she is working as a post-
doctoralfellowintheUniversityofManitoba,Canada.Herresearchinterests
includewirelesscommunicationswithfocusoninterferencemodeling,spec-
trumallocation,andpowercontrolinheterogeneousnetworks.
Ekram Hossain (S’98-M’01-SM’06) is a Profes-
sor in the Department of Electrical and Com-
puter Engineering at University of Manitoba, Win-
nipeg, Canada. He received his Ph.D. in Elec-
PLACE trical Engineering from University of Victoria,
PHOTO Canada, in 2001. Dr. Hossain’s current research
HERE interests include design, analysis, and optimization
of wireless/mobile communications networks, cog-
nitive radio systems, and network economics. He
has authored/edited several books in these areas
(http://home.cc.umanitoba.ca/∼hossaina). Dr. Hos-
sainservesastheEditor-in-ChieffortheIEEECommunicationsSurveysand
Tutorials and an Editor for IEEE Journal on Selected Areas in Communi-
cations - Cognitive Radio Series and IEEE Wireless Communications. Also,
he is a member of the IEEE Press Editorial Board. Previously, he served
as the Area Editor for the IEEE Transactions on Wireless Communications
in the area of “Resource Management and Multiple Access” from 2009-
2011 and an Editor for the IEEE Transactions on Mobile Computing from
2007-2012. He is also a member of the IEEE Press Editorial Board. Dr.
HossainhaswonseveralresearchawardsincludingtheUniversityofManitoba
MeritAwardin2010(forResearchandScholarlyActivities),the2011IEEE
Communications Society Fred Ellersick Prize Paper Award, and the IEEE
WirelessCommunicationsandNetworkingConference2012(WCNC’12)Best
Paper Award. He is a Distinguished Lecturer of the IEEE Communications
Society(2012-2015).Dr.HossainisaregisteredProfessionalEngineerinthe Amr Abdelnasser (S’12) received his B.Sc. and
provinceofManitoba,Canada. M.Sc. degrees both in Electrical Engineering from
Ain Shams University, Cairo, Egypt, in 2006 and
2011, respectively. Currently, he is a Ph.D. student
PLACE intheDepartmentofElectricalandComputerEngi-
PHOTO neering,UniversityofManitoba,Canada.Hiscurrent
HERE research interests include interference management
and resource allocation in heterogeneous cellular
networks. He has served as a reviewer for several
majorIEEEconferencesandjournals.
Mehdi Rasti (S’08-M’11) received his B.Sc. de-
gree from Shiraz University, Shiraz, Iran, and the
M.Sc.andPh.D.degreesbothfromTarbiatModares
University, Tehran, Iran, all in Electrical Engineer-
PLACE ing in 2001, 2003 and 2009, respectively. From
PHOTO November2007toNovember2008,hewasavisiting
HERE researcherattheWireless@KTH,RoyalInstituteof
Technology, Stockholm, Sweden. From September
2010 to July 2012 he was with Shiraz University
of Technology, Shiraz, Iran, after when he joined
the Department of Computer Engineering and In-
formation Technology, Amirkabir University of Technology, Tehran, Iran,
where he is an Assistant Professor. From June 2013 to December 2013, he
wasapostdoctoralresearcherintheDepartmentofElectricalandComputer
Engineering, University of Manitoba, Winnipeg, MB, Canada. His current
researchinterestsincluderadioresourceallocationinwirelessnetworksand
networksecurity.
10
TABLEII
COMPARISONAMONGVARIOUSPOWERCONTROLOPTIMIZATIONPROBLEMSANDTHEIRCORRESPONDINGDISTRIBUTEDSOLUTIONSFORSINGLE-TIER
NETWORKS,ALONGWITHTHEIRAPPLICATIONANDGENERALIZATIONFORDESIGNINGCAPCSCHEMESIN5GNETWORKS
Power control in single-tier Power control in prioritized Joint cell association and
networks multi-tiernetworks power control in multi-tier
networks
Optimizationproblem(P1) With fixed BS assignment, With fixed BS assignment, Minimize aggregate power
minimize aggregate power minimize aggregate power subjecttominimum(different)
subjecttominimumtarget-SIR subjecttominimum(different) target-SIRs for users in
forallusers target-SIRs for users in different tiers and obtain BS
differenttiers assignment
Distributedsolutions TPC[8] TPC[8] Minimum effective
interference based cell
association+TPC[14]
Optimizationproblem(P2) Minimize outage ratio of all Minimize outage ratio of Minimize outage ratio of all
users LPUEs subject to zero-outage usersindifferenttiersandob-
forHPUEs tainBSassignment
Distributedsolutions TPC-GR[11],[9],[10] Openproblem Openproblem
Optimizationproblem(P3) Maximize aggregate through- Maximize aggregate through- Maximize aggregate through-
putofallusers putofalluserssubjecttozero- putofallusersandobtainBS
outageforHPUEs assignment
Distributedsolutions OPC[12] Openproblem Openproblem
Optimizationproblem(P4) Maximize aggregate through- Maximize aggregate through- Maximize aggregate through-
putofalluserssubjecttomin- putofLPUEssubjecttomini- putofalluserssubjecttomin-
imumtarget-SIRforallusers mumtarget-SIRforallusers imum target-SIR for all users
andobtainBSassignment
Distributedsolutions DTPC[13] Openproblem Minimum effective-
interference-based cell
association+HPC[15]